Kling Avatar 2.0 Standard API

Kling Avatar 2.0 Standard

Kling Avatar 2.0 Standard is built for production teams that need reliable talking-avatar video from one portrait and one driving audio track. Through multimodal grounding and cross-attention audio-visual alignment, it converts speech into precise lip motion, natural micro-expressions, and coherent head dynamics while keeping the source face recognizable. Optional performance prompts let you steer emotion, gesture rhythm, and speaking stance without changing the simple request contract. At 7 credits per second with ceil-second billing, it is the cost-efficient tier for ads, courses, localized spokespeople, and short-form creator workflows that prioritize lip-sync quality and identity stability.

Why Choose This?

• Cost-efficient talking-avatar generationBills at 7 credits per second based on server-detected audio duration rounded up, so teams can iterate digital-human clips without jumping to the Pro tier.

• Precise lip-sync from driving audioAligns mouth shapes to speech timing across clean voiceovers, including fast dialogue and multilingual deliveries, for watchable talking-head results.

• Natural facial micro-expressionsAdds subtle eye, brow, and cheek motion that follows speech energy, so the avatar feels alive instead of locked in a static portrait pose.

• Single-portrait identity anchoringDrives the performance from exactly one publicly accessible character image, keeping facial features and framing stable through the generated take.

• Prompt-guided performance controlOptional prompts steer emotion, gesture cues, and speaking stance while the audio remains the primary timing source.

• Multilingual voiceover readinessWorks with English, Chinese, Japanese, and Korean voice tracks so the same portrait can power localized product demos and outbound campaigns.

Parameters

How to use

1. 01

   Prepare a clear character portraitUse one well-lit front-facing portrait with a visible face, balanced exposure, and a clean background so identity and framing stay stable.

2. 02

   Provide 2–60 seconds of speech audioUpload a publicly accessible voice track between 2 and 60 seconds (up to 5 MiB). Prefer clean volume and low noise for accurate lip alignment.

3. 03

   Optionally add performance promptsDescribe one primary emotion or gesture direction—such as a friendly product demo stance—so the model can guide micro-expressions and body cues.

4. 04

   Submit generation and retrieve the videoSubmit the task, poll with the returned task_id until finished, then download the talking-avatar video from data.files[].file_url.

Pricing

Final charge uses the server-detected audio duration, rounded up to a whole second.

Use cases

• Ecommerce ads and product demosTurn a brand spokesperson portrait plus scripted voiceover into short product explainers with synchronized lip motion and natural speaking posture.

• Online education and course narrationReuse one instructor portrait across lesson audio to produce consistent talking-head segments for courses, onboarding, and training libraries.

• Creator short-form videoGenerate portrait-driven talking clips for Reels, Shorts, and TikTok-style channels when you need fast lip-synced delivery from existing voiceovers.

• Cross-language go-to-market messagingPair the same face with localized audio tracks to ship multilingual brand messages without reshooting a human presenter for every market.

• Virtual brand spokespersonBuild a reusable digital face for announcements, FAQ explainers, and campaign intros that keep identity consistent across multiple scripts.

Tips

• Choose a sharp, evenly lit portrait with a centered face and modest headroom so identity and framing remain stable through speech motion.
• Use denoised audio with steady volume and clear consonants; cleaner voice tracks improve lip alignment on fast or multilingual lines.
• Keep each prompt focused on one primary emotion or gesture so performance cues stay coherent with the audio timing.
• For longer scripts, plan narrations in 2–60 second segments and stitch finished clips in edit so each take stays within the audio window.

Notes

• Final billing uses the server-detected audio duration rounded up to a whole second, then multiplies by 7 credits per second.
• audio_url must be a publicly accessible file lasting 2–60 seconds and at most 5 MiB.
• image_urls accepts exactly one character portrait as a public HTTP(S) URL, image Data URI, or raw Base64 for each generation task.
• Output is a fully pre-rendered talking-avatar video clip designed for publishing and editing workflows.